Free
$0 always
Create, use, and share public skills and Binders. Includes 2 AI drafts a month.
Start freeA package manager for AI agent skills
Your AI instructions, versioned like code.
Version controlled, CLI accessible, and loaded just in time. Your skills live in Binders, and your agent takes one out at the moment the task calls for it — so a turn carries the instructions it needs instead of everything you own. Shorter turns, lower token costs.
Share a Binder privately with your team, or publish it for anyone.
Create and use your skills and Binders for free.
In one line: Bindry is a package manager for AI agent skills — version controlled, CLI accessible, and loaded just in time. Write your team's instructions once — brand voice, sales playbooks, email tone, writing style or coding standards — as reusable skills, group them into Binders, and your agent takes out only the ones a task calls for: in Claude Code, Codex and GitHub Copilot through plugins, or in any other AI tool through AGENTS.md, Markdown or MCP.
Use cases
Not just for code.
Anything you ask AI to do goes better with clear instructions. Write them once, and every AI tool you use follows them.
-
Marketing
Your brand voice, audience, and campaign rules, followed in every AI draft.
-
Sales
Your pitch, pricing rules, and answers to common objections, the same every time.
-
Email
Tone, sign-offs, and what never to promise, for every email AI writes for you.
-
Writing
Your style guide and words to avoid, for blog posts, docs, and social posts.
-
Code
Architecture rules, git flow, and review standards your coding agents follow.
How it compares
Against what you are doing today.
Most teams keep their AI instructions in a folder of files, or paste them in each time. Both work until the instructions matter.
Versioned, and pinned where you installed it
Every skill carries a version. An install pins one, and tells you when it has moved on.
- Bindry
- Yes
- A folder of SKILL.md files
- Partly
- Prompts pasted into chat
- No
Loads only what the task needs
Each skill compiles to its own triggered file. A ten-skill Binder does not cost ten skills a turn.
- Bindry
- Yes
- A folder of SKILL.md files
- Partly
- Prompts pasted into chat
- No
Says when it does not apply
A rule with no stated limits is one the agent applies where it does harm.
- Bindry
- Yes
- A folder of SKILL.md files
- No
- Prompts pasted into chat
- No
Says how to tell it held
Verification steps travel with the skill and compile into the file the agent reads.
- Bindry
- Yes
- A folder of SKILL.md files
- No
- Prompts pasted into chat
- No
Shows the cost before you install
A token estimate on every skill, and on the Binder as a whole.
- Bindry
- Yes
- A folder of SKILL.md files
- No
- Prompts pasted into chat
- No
Keeps always-on voice apart from triggered skills
Tone and house style ride along with the Binder instead of being forced into a trigger.
- Bindry
- Yes
- A folder of SKILL.md files
- No
- Prompts pasted into chat
- No
Private to your team
Shared inside your workspace, visible to the people you work with and nobody else.
- Bindry
- Yes
- A folder of SKILL.md files
- Partly
- Prompts pasted into chat
- No
Anyone can author one
A guided editor and AI-assisted drafting, instead of frontmatter you write by hand.
- Bindry
- Yes
- A folder of SKILL.md files
- No
- Prompts pasted into chat
- No
Partly means it is possible with discipline and tooling of your own — a repository you remember to tag, a host that happens to load files on demand, a private repo — rather than something the format gives you.
How it works
A Binder isn't one giant prompt copied into every request — it only loads the exact instructions each task needs.
- 1
Compose
Pick the skills your project needs and put them in order to build a Binder.
- 2
Check the cost
See how much space it takes up before it's added to every request.
- 3
Export
Turn it into a prompt, a document, or a ready-made file your AI tool can read.
binder.pull()
$ bindry pull git-flow-command-center
resolves: branch strategy, commit rules, PR review, CI triage
writes: one SKILL.md per skill — the format Claude Code and Codex already read
saves repeated context tokens every sprintDon't see a skill for what your team needs yet? A Binder is only as good as the skills in it — write your own and it's ready to add to a Binder like this one.
Prefer a terminal? npx bindry search browses the Library with no account needed.
Works where you already use AI
Install a plugin and your Binders show up inside the AI tool you already use. No copying and pasting.
Claude Code
Turns a Binder into skills Claude loads only when a task needs them.
OpenAI Codex
The same plugin, built for Codex.
GitHub Copilot
The same plugin, built for Copilot. New, in preview.
Any other AI tool
Nothing to install: use an AGENTS.md file, a Markdown bundle, or MCP.
Why Bindry
Agent instructions became a package ecosystem. Most of it still ships like loose files.
Every team using AI to build has a folder of instructions that make their agent behave — architecture rules, review standards, deployment steps, brand voice. Most of those files still get shared the way they always have: somebody copies them. No version. No provenance. No way to know the copy still matches the tool it was written for.
This isn't a hunch. It's been measured.
GitSkills · MSR 2027 · July 2026
3,797,117
skill files found in public repositories
282,200
repositories carrying them, across 195,841 accounts
50.5%
are verbatim copies of another file
0
registries tracking any of it
Four university researchers catalogued the entire public population of agent instruction files and published it as an open dataset. Their own framing for what they found: reuse "without a package manager," in a "registry-free ecosystem" where a modified copy can quietly introduce command execution the original never had. arXiv:2608.10906 →
Storing the file was never the hard part.
Instruction files are easy to put somewhere. GitHub already holds millions of them. What no repository can tell you is the part that actually matters once more than one person depends on them: which version is running, what it composes with, what it costs you on every request, and whether anyone approved the change.
That gap is not a storage problem. It's the difference between a pile of files and a system — and it's the entire reason Bindry exists.
What Bindry adds
Four things a folder of files can't do.
A skill is one instruction with a purpose and a version. A Binder is an ordered, pinned set of them, compiled for the AI tool you actually use.
Composition, not collection
Pick the skills a project needs, set their order, and pin the versions. A Binder compiles to one coherent instruction set instead of a directory you hope your agent reads correctly. When a standard changes badly, roll it back.
Ordered · versioned · pinned · reversible
Context cost, before you ship it
Instructions aren't free. They're re-sent on every request, and a bloated Binder quietly taxes every task your agent runs. Bindry shows you what a skill and a compiled Binder cost in context, so you can keep guidance tight instead of paying for a wall of text forever.
The only line item nobody else on your binder is measuring.
Only loads what a task needs
Every skill in a Binder compiles to its own skill file with its own trigger, not one merged block of text. Add a tenth or a twentieth skill for an edge case your team hits twice a year, and it costs nothing on the other 363 days — only the skills a task actually touches load for it.
Per-Skill triggers · Binder size ≠ per-task cost
One model, every agent surface
Bindry keeps one canonical version of your standards and compiles it out to the tools your team already uses. You write the rule once. It shows up correctly wherever the work happens, and it survives your team changing tools next quarter.
Write once → Claude Code · Codex · AGENTS.md · MCP
New to AI? Start here.
A plain-language walkthrough of how AI actually works, and why your instructions keep disappearing. No coding required.
Free to start
Upgrade when you want privacy or a team.
Pro
$10 per month
Create private skills and Binders, pin versions, and reuse expensive build context.
Sign up for ProTeam
$18 per seat, per month
One private library your whole team pulls from, pinned so everyone works from the same version.
Sign up for TeamAll prices are in US dollars (USD). If your card is from outside the US, checkout may show the price in your local currency, converted by Stripe when you pay. The amount shown at checkout is what you are charged for that payment.
Public
Shows up in the Library for anyone to find, copy, and learn from. Good for sharing what you know.
Private
Visible only to you and your workspace — nothing leaves your team. Requires Pro or Team.
Teach your AI once. Use it everywhere.
Start with a Binder from the Library, or write your own in a few minutes.